Design of Optimal Deep Learning Based Human Activity Recognition on Sensor Enabled Internet of Things Environment

In recent times, Human Activity Recognition (HAR) has become a major challenge to overcome among computer vision applications in day-to-day lives. HAR is mainly envisioned to be utilized in coordination with other technologies namely, Internet of Things (IoT) and sensor technologies. Due to the advancements made in Deep Learning (DL) approaches, the automated high level feature extraction process can be utilized to improve the outcomes of HAR process. In addition, DL techniques can also be employed in different domains of sensor-enabled HAR. In this aspect, the current study designs an Optimal DL-based HAR (ODL-HAR) model on sensor-enabled IoT environments. The proposed ODL-HAR technique aims at determining the human activities in day-to-day lives using wearables and IoT devices. ODL-HAR technique involves different stages of operations namely, data acquisition, data preprocessing, feature extraction, classification, and parameter optimization. The proposed ODL-HAR technique uses MobileNet-v2 model as a feature extractor and Bidirectional Long Short-Term Memory (BiLSTM) model as a classifier. In order to fine tune the hyperparameters involved in BiLSTM model optimally, Chaos Game Optimization (CGO) algorithm is employed which in turn increases the recognition performance. The novelty of the work lies in the deployment of CGO algorithm for hyperparameter optimization of HAR. A wide range of simulations was conducted to validate the supremacy of the proposed ODL-HAR technique and two benchmark datasets were used for this simulation process. The experimental results portrayed the enhanced performance of ODL-HAR technique over other recent HAR approaches under different evaluation parameters.